PROJECT SUMMARY BACKGROUND: Depression and anxiety are the leading causes of disability and lost productivity, and are often underdiagnosed and undertreated owing to access, cost, and stigma barriers. Novel and scalable psychotherapies are urgently needed. Advances in artificial intelligence (AI) offer a transformative opportunity to develop intelligent voice assistants as virtual health agents accessible on personal devices. Meanwhile, major advances in human neuroscience have fueled a paradigm shift to study brain mechanisms underlying behavioral health interventions. OBJECTIVES: Leveraging our collaborative team?s transdisciplinary expertise in these emerging areas, we will develop and rigorously test a novel voice-enabled, AI virtual agent named Lumen, trained on Problem Solving Therapy (PST), for patients with moderate, untreated depressive and/or anxiety symptoms. We will investigate the effect of Lumen on engagement of a priori neural targets?amygdala for emotional reactivity and dorsal lateral prefrontal cortex (DLPFC) for cognitive control?as putative mechanisms. DESIGN/ METHODS: The project has 2 phases. In the R61 phase (years 1-2), we will further develop Lumen building on the current prototype and conduct iterative user-centered design evaluations that include focus groups, scenario-based clinician evaluations, and a formative user study with 20 participants. We will pilot test Lumen in a 2-arm randomized clinical trial (RCT, Study 1), with 60 participants with depression and/or anxiety randomized in a 2:1 ratio to receive PST with Lumen (n=40) on a secure study iPad or be on a waitlist (n=20). At weeks 0 and 14, participants will complete functional magnetic resonance imaging (fMRI) to assess neural target engagement as well as validated surveys of patient-reported outcomes (e.g., depressive and anxiety symptoms, functioning, quality of life). In addition, they will complete naturalistic end-of-day assessments of mood, stress, appraisal and coping for 7 days every 2 weeks. If the Go milestone criteria are met, the R33 phase (years 3-5) will include a 3-arm RCT (Study 2) with 200 new participants randomized in a 2:1:1 ratio to 1 of 3 arms: Lumen (n=100), waitlist control (n=50), and in-person PST as active control (n=50). Participants will complete baseline and follow-up assessments using a refined measurement protocol based on Study 1. SPECIFIC AIMS: R61 aims are to (1) establish the functionality, usability, and treatment fidelity of Lumen; and (2) demonstrate feasibility, acceptability, and neural target engagement according to pre-specified Go milestone criteria. R33 aims are to (1) confirm neural target engagement by a superiority test (primary) comparing the Lumen and waitlist control arms and a noninferiority test (secondary) comparing the Lumen and in-person PST arms; and (2) examine the relationships of target engagement to outcomes. The results will provide the basis for future confirmatory efficacy testing of Lumen. IMPACT: This project?s public health impact lies in that a mechanistically tested, PST-trained AI agent could bring proven psychotherapy to people with depression/anxiety who do not seek professional help or who desire more personalized, connected care.